Recognition Of Mathematical Notation

This paper provides a compelling illustration of the complexity of the mathematics-recognition task, of the many symbol configurations that must be considered. The authors state that syntactic approaches, using parsing, are untenable because the great variety of possible expressions makes it impossible to provide an a priori syntax definition for all possible expressions. This, combined with the computational complexity of parsing, motivates them to instead use a large collection of procedurally-coded recognition rules. Their comment about a priori syntax definition requires some clarification: any recognition method, including procedurally-coded rules, implicitly or explicitly defines the syntax of recognizable expressions. Apparently Okamoto et al. find it easier and more practical to provide an implicit syntax definition, coded as procedural rules, rather than an explicit syntax definition, coded as grammar rules